A comparative study on weighting-based clustering techniques: Time series data

Abstract

Lately, it seems that has been increased trend in time series data mining in different applied domains. Time series data is an ordered sequence of real values into specified temporal intervals. Mining of time series data follow out four fundamental goals that clustering is one of the important goals. Time series clustering is to discover patterns of similar time series in a database of different sciences. So far, various techniques have been proposed in the field of time series clustering that technique of weighting-based clustering has been used widely in the last dedicate researches. Weighting-based clustering is a technique based on feature-weighting that with assignment weights to features can indicate importance feature and improve performance time series clustering. Indeed, feature-weighting can be considered as a feature extraction technique that it has many applications in high dimensional data such as time series. But, better results provide with regards to assignment weights to features compared to other feature extraction techniques. Also, it can be used as a useful technique to reduce dimensions and improve performance clustering task in high dimensional data. Hence, this paper performs a comparative study on weighting-based clustering techniques on time series data. For this goal, at first, the paper tries to explore related researches. Then, provide an extensive evaluation and comparison of weighting-based techniques for clustering of time series data. The behavior of weighting-based techniques has been reported for clustering of three well-known time series data set. Experiments show the improvement of performance of clustering by using the idea of feature-weighting based on optimization. In the other word, feature-weighting based on optimization can be useful for helping an appropriate decision towards selecting weighting-based clustering technique for researchers' needs.